Uncertainty quantification for oxidative kinetic parameters in smoldering model with oxygen nonequilibrium concept

Abstract

Uncertainty quantification is very crucial for combustion models. Smoldering models have been questioned by large uncertainties resulting from the underlying multiphase and multiscale physics. The confidence for most of smoldering models has been unknown because their uncertainties have been rarely quantified. This work attempts to quantify the uncertainties of oxidative kinetic parameters in a newly developed smoldering model with oxygen nonequilibrium concept that has been demonstrated to successfully predict smoldering under a variety of conditions covering from propagation to extinction as well as from buoyancy-driven fires to applied smoldering engineering. The uncertainty quantification is comprehensively investigated by 135 models in total, which involves 15 sets of oxidative kinetic parameters with the aleatory uncertainty (Φ_oxid) of TG-scale experiments ranging between 0.04 and 0.6 and three physic variables, i.e. Darcy flux (2.7 cm s^-1 – 21.2 cm s^-1), fuel type (PSFs including bituminous coal and anthracite and CFIPM i.e., food waste in sand), and spread mode (forced reverse, forced forward and buoyancy-driven forward). Results show that the confidence for smoldering model with oxygen nonequilibrium concept is rather good with the model bias less than 0.25 if the aleatory uncertainty is smaller than 0.15. Particularly, the confidence is extremely good for forced reverse smoldering of bituminous coal since the model bias is always less than 0.25 even though the oxidative kinetic parameters’ uncertainty reaches as high as 0.57. Besides, the model confidence for CFIPM is better than PSFs as the fuel loads become less. With the oxygen nonequilibrium concept, the apparent kinetic parameters with Φ_oxid < 0.15 could achieve a rather low and acceptable bias of model prediction, which is beneficial to save enormous time and efforts by avoiding to seek the intrinsic kinetic parameters from a large high-dimensional space. Nevertheless, the model confidence decreases for large Darcy flux > 20 cm s^-1, forward spread, and buoyant effect. It is for the first time that the uncertainty quantification for input oxidative kinetic parameters optimized from TG-scale experiments is comprehensively investigated for smoldering models. This work elucidates the confidence of smoldering models with oxygen nonequilibrium concept and improves our understanding how the aleatory uncertainty induced by kinetic parameters’ optimization propagate in multiscale modelling for smoldering combustion.